Back to portfolio

AI Media Automation Portal

A Human-in-the-Loop AI platform that automates content discovery, script generation, video creation, YouTube publishing, and analytics.

Agentic WorkflowsHuman-in-the-LoopLovableSupabasen8nYouTube APIAI ScoringScript GenerationVideo Automation

Media Automation · Studio

Pipeline running

Video timeline

Awaiting approval Approve
Project Overview

AI Media Automation Portal is an end-to-end AI-powered content automation platform designed to streamline the full media creation lifecycle — from content discovery and script generation to video creation, human approval, YouTube publishing, and performance tracking.

Business Problem

Manual content creation is repetitive

Manual content creation requires multiple repetitive steps:

1

Searching for trending topics

2

Reviewing multiple websites

3

Selecting the best topic

4

Writing YouTube-friendly scripts

5

Creating voiceover

6

Finding visuals and B-roll

7

Rendering video

8

Uploading to YouTube

9

Tracking performance metrics

The platform solves this by creating a configurable AI workflow where content generation can be automated, reviewed, and published from one centralized dashboard.

Key Objectives

What the platform delivers

Automate topic-based content discovery

Region-specific selection: Global, Middle East, South Asia, UAE, India

Generate AI-ranked top content ideas

Enable Human-in-the-Loop dashboard approvals

Generate short-form YouTube scripts

Support script regeneration and AI-assisted improvement

Generate video assets and final video output

Upload approved videos to YouTube

Track views, likes, comments, and engagement

Maintain dropdowns and sources from backend master tables

My Role

Solution Lead & Architect

Nikhil designed the overall solution architecture, user journey, backend data model, automation orchestration approach, Human-in-the-Loop approval model, Supabase backend structure, LOV-based master data design, Lovable frontend pages, n8n workflow orchestration, AI prompts, YouTube upload process, and scalable multi-topic / multi-region architecture.

Technology Stack

The tools behind the platform

Frontend

Lovable

Backend Database

Supabase PostgreSQL

Authentication

Supabase Auth

Storage

Supabase Storage

Workflow Automation

n8n

AI Orchestration

n8n AI workflows

AI Capabilities

Prompt engineering, scoring, scripts, refinement

Content Sources

RSS feeds, websites, scraping sources

Video Workflow

AI voice, captions, B-roll, rendering

Publishing

YouTube Data API

Analytics

YouTube Analytics / Data API

Governance

Human-in-the-Loop checkpoints

High-Level Architecture

Three major layers

01

Frontend Layer

Lovable

  • Login page
  • Setup page
  • Ideas dashboard
  • Script review page
  • Video preview page
  • YouTube metrics dashboard
  • Admin master data page
02

Backend Layer

Supabase

  • User authentication
  • Master data
  • LOV dropdown values
  • Topic and region website sources
  • User configurations
  • Generated content ideas
  • Scripts and script versions
  • Video records
  • YouTube upload records
  • Metrics
  • Workflow logs
03

Automation Layer

n8n

  • Scheduled execution
  • Manual Run Now
  • Website scraping
  • AI content scoring
  • Top 3 idea generation
  • Script generation
  • Script regeneration
  • Script improvement
  • Video generation
  • YouTube upload
  • Metrics sync
  • Workflow logging
Configurable Setup

One dashboard to configure the pipeline

Content Automation SetupDraft

Topic

AI & Technology

Region

Middle East

Frequency

Daily

Generation Time

08:00 GST

Language

English

Tone

Informative

Video Duration

60 seconds

YouTube Visibility

Public

Website Sources

8 selected
CancelSave Configuration
LOV-Based Master Data

Backend-driven dropdowns

All dropdowns are maintained in Supabase backend tables and are not hardcoded in the frontend. This makes the platform flexible and scalable.

lov_topics
lov_regions
lov_frequencies
lov_languages
lov_tones
lov_video_durations
lov_youtube_visibility
topic_region_sources
Human-in-the-Loop Approval Flow

Approval checkpoints at every stage

01

Idea Review

Approve one idea or reject all

02

Script Review

Approve, regenerate, or edit and improve

03

Video Preview

Approve upload or request changes

04

Publishing

Upload approved video to YouTube

Workflow Architecture

Orchestrated workflow nodes

Generate Topic Sources

01

Run Now

02

Scheduled Runner

03

Scrape and Score Ideas

04

Approve Idea

05

Reject All Ideas

06

Generate Script

07

Regenerate Script

08

Improve Script

09

Approve Script

10

Generate Video

11

Approve Upload

12

Sync Metrics

13
Example End-to-End Flow

From login to live metrics

1

User logs into Lovable portal

2

User selects topic and region

3

Supabase returns website sources

4

User saves configuration

5

n8n runs workflow

6

n8n scrapes selected websites

7

AI scores articles

8

Top 3 ideas are stored in Supabase

9

User approves one idea

10

AI generates YouTube script

11

User reviews script

12

AI generates video

13

User previews video

14

User approves upload

15

n8n uploads to YouTube

16

Metrics sync back to Supabase

Key AI Concepts

The applied AI stack

Agentic Workflow Orchestration

Coordinated AI steps run discovery, scoring, scripting, and publishing across n8n.

Prompt Engineering

Structured prompts control scoring criteria, script tone, and output formatting.

AI-Based Content Scoring

Models rank scraped articles by relevance and engagement potential.

Generative AI Script Creation

Short-form YouTube scripts generated from the approved idea.

Human-in-the-Loop Governance

Approval checkpoints at every critical decision before publishing.

AI-Assisted Content Refinement

Scripts can be regenerated and improved with AI suggestions.

Performance Feedback Loop

YouTube metrics loop back into Supabase to inform future scoring.

Benefits Delivered

Measurable outcomes

Reduced manual effort in content research

Faster content idea discovery

Consistent script quality

Better governance through approval checkpoints

Centralized content operations dashboard

Scalable topic and region configuration

Reusable AI workflow architecture

Better visibility into YouTube performance

Project Impact

This project demonstrates how AI can be used beyond simple chatbot use cases and applied to a real operational workflow — combining generative AI, agentic automation, Human-in-the-Loop governance, workflow orchestration, backend-driven configuration, media automation, and performance analytics.

Explore how AI can automate the full content operations lifecycle